Is quantum computing here? A researcher's reality check

by Daniel Cohen

Short answer: useful quantum advantage exists in a handful of narrow, mostly contrived problems. Practical, error-corrected machines that break RSA or design drugs are still years out — error correction is the real mountain, not qubit count. The press conflates "we made N qubits" with "we solved your problem". We didn't. That said, progress on logical qubits in the last two years is genuinely encouraging. Don't sell your crypto over it yet, but don't sleep on it either.

favorite 50 comment 5 visibility 1287

Comments

Lena Novak 3 months ago

This. Logical qubits are the metric that matters and almost no headline reports them.

Priya Nair 3 months ago

Appreciate a researcher saying the quiet part: most of it is still contrived benchmarks.

Lena Novak 1 month ago

Materials person seconding the reality check: half of the "quantum progress" headlines are actually materials-science progress wearing a costume. The qubits improved because the substrates and shielding improved. The unsung heroes have lab coats and furnaces.

Jonas Iversen 1 month ago

Seconding Lena from the next bench over, one layer down: even error correction is a materials and refrigeration problem. The wall isn't "more qubits," it's keeping a few thousand of them cold, clean and coherent at the same time — that's fabrication yield and thermal budget, not algorithms. Logical qubits get better when the fridge and the fab do.

Ava Engel 1 month ago

Late here, but it's my field: Daniel's reality check has aged well enough to earn a footnote. Lena and Jonas are right that the fridge and the fab set the ceiling. But what actually moved this year is classical: real-time decoding dropped to sub-microsecond. A decoder that can't keep pace with the syndrome stream isn't error correction, it's logging, and that was quietly the binding constraint on half the surface-code demos people cite at me. Second, Zuchongzhi 3.2 got below threshold on all-microwave control - a different pathway, which matters more than its qubit count. One system below threshold is a result; two by different routes is a trend. Still nowhere near RSA. Watch logical error rates and decoder latency; close the tab if paragraph one says Bitcoin.